Abstract: From a knowledge perspective, a product constitutes a knowledge set. For a long time, academic research has conflated knowledge production in the product R&D phase with that in the product usage phase. Focusing on the issue of knowledge production during new product usage, this paper selects three cases—Comet airliner, lithography machine, and intelligent driving vehicles—and through comparative research, finds that: First, from a temporal dimension, the knowledge set of a new product comprises laboratory knowledge from the R&D phase and user knowledge from the usage phase. User knowledge, historically overlooked, refers to new knowledge generated during the usage phase that extends beyond the scope of laboratory R&D. This new knowledge arises from unforeseen actual usage scenarios and the discrepancies between designed and real-world scenarios. The production of user knowledge embodies the technological iteration of new products: in the iterative process during the usage phase, each instance of user knowledge produced serves as a knowledge reference for the next round of laboratory-based product R&D. User knowledge and laboratory knowledge are non-substitutable and complementary. Second, unlike laboratory knowledge production—where both “problem identification” and “problem solving” are undertaken by product manufacturers—user knowledge production follows the technological iteration logic of lead users raise problems—manufacturers solve problems. As new products may expose multiple issues during use, a small subset of users (i.e., lead users) bear the uncertainties, risks, and costs associated with new product adoption. Once a new product completes user knowledge production, it matures, and the broad base of ordinary users then becomes the foundation supporting the evolution of the new product into a new industry. Third, generative artificial intelligence (AI) has transformed the mode of user knowledge production. Manufacturers and users can now synchronously perceive the usage environment and user behavior, enabling manufacturers to achieve vertically integrated user knowledge production. Generative AI narrows and blurs the boundary between ordinary users and lead users by “injecting the wisdom of predecessors into users’ minds,” reducing users’ uncertainties, risks, and costs, and empowering ordinary users to become lead users involved in user knowledge production. This assists manufacturers in producing user knowledge more rapidly and with higher quality, generating “intelligent solutions” for technological iteration. Based on these findings, three implications are proposed. Firstly, a lead user supply mechanism must be established. Without lead users’ environmental perception and problem identification, manufacturers cannot complete technological iteration of new products, thereby impeding the formation of new development trajectories and industries. Secondly, the development of generative AI must be accelerated. Generative AI-empowered new products are capable of perceiving usage scenarios and collecting user behavior data more comprehensively, accurately, and rapidly. This capability enables broader participation of ordinary users in user knowledge production while promoting exponential improvements in corporate efficiency for user knowledge generation. Thirdly, the data asset management system must be rapidly improved. User knowledge production for new products—whether or not generative AI is employed—involves the use of user data. Accelerating the improvement of user data legislation and establishing a sound data property rights system will facilitate the advancement of user knowledge production.